Successes and challenges of near-infrared spectroscopy in the

Actas • Proccedings POSTCOSECHA E INDUSTRIALIZACIÓN • POSTHARVEST AND PROCESSING

Successes and challenges of near-infrared spectroscopy in the avocado value Chain R.J. Blakey1, Z. Van Rooyen 1, J. Berry 2, M. Elliott 3, S. Rusby 4 . Westfalia Technological Services, Tzaneen, South Africa

1

. Greencell, Spalding, United Kingdom

2

. Westfalia Fruit Estates, Tzaneen, South Africa

3

. Taste Technologies, Auckland, New Zealand

4

ABSTRACT Two near-infrared spectrometers (NIR) were tested at various points in the avocado value chain. The first was a handheld unit for estimating dry matter content. When measurements were taken with the fruit skin removed, the R2 = 83% and SEP = 2.25% DM for ‘Fuerte’, ‘Hass’, and ‘Mendez #1’ (‘Carmen®-Hass’) combined. A comparison between the conventional method to determine dry matter and this NIR method, showed that, on an orchard basis, 76% of the DM-values being within 1.5% DM of each other. After a robust model has been developed, handheld NIR can be used to quickly determine dry matter of avocados pre- and postharvest. The second unit was a packline-mounted T1 unit from Taste Technologies for application in a packhouse or ripening facility. On hard fruit, the unit was able to eliminate class 1 fruit with orchard cold (frost) damage with accuracy of up to 90%, depending on severity of the damage. This unit was used on a semi-commercial basis at Westfalia Packhouse in 2014. Furthermore, it was able to estimate dry matter with a standard error of prediction (SEP) of 1.9% DM and R2 = 76% - comparable to the handheld unit. For ripe fruit, the T1 was able to detect diffuse flesh discolouration (grey pulp) with an accuracy of about 80%, depending on severity. Other disorders and diseases were not consistently detected due to the isolated nature of these defects in the fruit flesh. The economic feasibility of online NIR is complex and needs to be determined case by case. Keywords: Maturity, Dry matter, Non-destructive testing, Fruit quality, Frost damage, Grey pulp.

INTRODUCTION Non-destructive testing of avocado maturity and fruit quality using near-infrared spectroscopy (NIR) has been investigated by various researchers (Schmilovitch et al., 2001; Clark et al., 2003; Blakey et al., 2007; Blakey et al., 2009; Blakey & van Rooyen, 2011; Blakey, 2012; Blakey, 2014). The online non-destructive determination of fruit maturity and quality is desired by packhouses and ripening centres (RCs) as a means to identify and reject immature and poor quality fruit from the supply chain at these two critical handling points, i.e. before packaging and shipping, and distribution to supermarkets, respectively. Another option is to use a portable (i.e. handheld) instrument to determine fruit maturity and quality at locations away from the packline. This can be throughout the supply chain. Fruit Maturity Early season avocado fruit fetch a price premium because demand exceeds supply. Growers are therefore eager to supply this market. In South Africa, this has resulted in some growers harvesting and supplying immature fruit. Immature fruit does not complete ripening and the quality of the fruit is poor. This resulted in the South African Avocado Growers’ Association (SAAGA) implementing a campaign against the supply of immature fruit (Muller, 2010). A portable means to determine fruit maturity would allow for the rapid testing and removal of immature fruit from the supply chain. Days to Ripen The variation in days to ripen results in increased costs for a RC, and fruit quality loss as fruit can be bruised after repeated handling. As such, RCs desire to reduce this variation in ripening. The ability to pre-sort fruit before ripening, preferably at the packhouse before export, would be highly advantageous. Fruit Quality A major avocado fruit quality disorder is grey pulp – also termed mesocarp discolouration or diffuse flesh discolouration (White et al., 2004). The incidence of the defect increases considerably later in the avocado season, and limits the export season in South Africa. Grey pulp develops as the fruit ripens, but the defect is strongly related to preharvest factors (van Rooyen & Bower, 2005) so it would be commercially valuable to sort fruit prone to grey pulp at the packhouse and/or at the RC. Avocado fruit are sensitive to orchard cold damage. The fruit flesh turns brown and the damaged tissue does not ripen. The fruit skin can turn red or bronze but there is not a strong correlation between skin and flesh symptoms. Therefore it is not possible to visually sort fruit visually with any degree of accuracy. A non-destructive method to detect frost damage offers a means to recover undamaged fruit.

VIII Congreso Mundial de la Palta 2015 | 410


Objectives 1. Calibrate a handheld NIR spectrometer for avocado moisture content (MC) and determine if, and where, this instrument is commercially viable. 2. Develop calibrations models for an online NIR spectrometer for avocado MC, internal defects, and days to ripen. Then determine if, and where, the instrument is commercially viable.

MATERIALS & METHODS Handheld NIR A Phazir 1018 handheld near-infrared spectrometer (Thermo Scientific, Wilmington, MA, USA) was used in this study. For the calibration of the instrument, fruit were scanned at four to six points around the circumference. Moisture content (MC; the complement of dry matter) was measured gravimetrically within a few hours of harvest; 1.0g samples were oven-dried at 70-75°C for at least 24 h (Blakey, 2013; Blakey, 2014). The model was developed between 2011 and 2013 and then validated in 2014 and 2015. About 10,000 samples were used in the final models, due to a limitation in the instrument’s software (Polychromix-MG v 3.101.0.0). Online NIR A T1 (Taste Technologies, Onehunga, New Zealand) NIR spectrometer was used. MC was measured according to the above protocol. For the calibrations for defects, fruit were scanned on the day of harvest, approximately 12h after removal from 28 d of cold storage (once the fruit temperature had reached room temperature), and at ripeness. Fruit were scanned twice each, with a 180° rotation between scans. Fruit were cut and assessed at ripeness according to White et al. (2004). Models were developed between 2010 and 2013 according to Blakey (2012). The model for grey pulp was validated in 2013 using ‘Fuerte’ that was stored at 5.5°C for 28 days.

RESULTS AND DISCUSSIONS Handheld NIR The MC of ‘Fuerte’ fruit could be estimated with an acceptable level of accuracy, but the accuracy of MC for ‘Hass’ and ‘Carmen®-Hass’ was poor if the skin was not removed. A model for MC with the fruit skin removed combining all three cultivars was developed. This is preferable because the operator would not have to change models when scanning different cultivars, and it may be useful for cultivars not included in the model. The external validation of the handheld NIR showed that the combined model for flesh MC with the skin removed is stable (Figure 1) with an error almost equal to the conventional gravimetric method used at Westfalia Fruit Packhouse (Table 1). The predictive ability for the model was lower for ‘Hass’, but this appears to be because the gravimetric MC was unexpectedly too low when compared to the previous week and results from neighbouring orchards. As with conventional methods to determine fruit maturity, some interpretation of the results is necessary to obtain logical results.

Figure 1: External validation of ‘Carmen®-Hass’, ‘Fuerte’, and ‘Hass’ in 2014 and 2015. Fruit flesh moisture content (MC) determined gravimetrically vs. handheld NIR (skin removed) where each data-point is for a Westfalia Fruit orchard at a specific date.

VIII Congreso Mundial de la Palta 2015 | 411


Table 1: Co-efficient of determination (R2) and standard deviation (SD) for fruit flesh moisture content (MC) determined gravimetrically and by handheld NIR in the external validation. SD was calculated per orchard per date Cultivar

R2

SD Gravimetric (% MC)

SD NIR (% MC)

Carmen

78%

1.8

1.9

Fuerte

78%

1.9

1.9

Hass

71%

2.1

2.0

Sources of Error There are a number of sources of error when calibrating and using the handheld NIR. Some of these are discussed briefly below. 1. Error in oven-drying method. This can be checked by including multiple samples from the same fruit so that erroneous values can be detected and removed. 2. Water stress in-field. Fruit from dryland orchards, or from water stressed trees, have shown high variability in the MC results (data not shown). 3. Temperature. The principle of NIR is based on the chemical bonds within molecules. These are affected by temperature, so large temperature variations will increase the error in NIR readings. Also, fruit with a high pulp temperature will likely be dehydrated. Cold storage and ripening. The texture and chemical composition of avocado fruit changes greatly during cold storage and ripening (Blakey et al., 2012; Blakey et al., 2014). Fruit MC should be measured as soon after harvest as possible and is a non-meaningful parameter after cold storage or once fruit begin to ripen. 4. Poor contact between instrument and sample. Care should be taken to have good contact to reduce “leak light” increasing noise in the spectra. Commercial Considerations The recommendation is to use the average fruit MC (sample size of at least 4 fruit and at least 2 measurements per fruit) rather than consider individual fruit. There is considerable natural variation between avocado fruit so it is advisable to maximise the sample size for better results. At Westfalia Fruit, pre-season maturity testing includes MC, but also includes samples to determine fruit quality upon ripening. The legal minimum MC and acceptable fruit quality have to be achieved before an orchard can be harvested. However, there is a bottleneck at the MC laboratory, due to the scale of the Westfalia Fruit operation. The handheld NIR can be used to ease this bottleneck by increasing the analysis speed. Three seasons were required to develop a stable and accurate model for this handheld NIR. This concurs with Wedding et al. (2011). The cost of a unit, a long lead-time, and specialised skills required to calibrate the instrument, are prohibitive to most companies wishing to make use of this technology. Online NIR Fruit Maturity The online NIR was able to measure the MC of avocado fruit with fairly good accuracy (Table 2). This is slightly more accurate than the handheld NIR, and also faster (